Figure 2 - uploaded by Dar A. Roberts
Content may be subject to copyright.
(a) Examples of crown objects overlaid on a 1 m lidar canopy height model. (b) The same crown objects overlaid on a 3.7 m colour infrared (R, 724 nm; G, 648 nm; B, 550 nm) display of AVIRIS imagery.
Source publication
In this research, we classify 15 common urban trees in downtown Santa Barbara, California, using crown-level canonical discriminant analysis (CDA) on airborne visible/infrared imaging spectrometer (AVIRIS) imagery. We compare the CDA classification accuracy against results obtained fromstepwise discriminant analysis. We also examine the impact of v...
Similar publications
The aim of this study was to discriminate the geographic origin of Korean, Chinese, and Indian sesame oils distributed in Korea using (1)H NMR spectroscopy in combination with canonical discriminant analysis (CDA). (1)H NMR spectra were obtained from 84 roasted oil samples prepared from 51 Korean, 19 Chinese, and 14 Indian sesame seeds. The integra...
Citations
... Remotely-sensed satellite imagery can be used to interpolate data from field plots based upon variability of spectral signatures, estimating variation and approximating species composition across landscapes (Adelabu et al., 2013), but the spatial and temporal resolution of most satellite imagery remains a constraint (Carleer and Wolff, 2004), and higher resolution data, such as those collected from aircraft, are needed to monitor individual trees (Bergseng et al., 2015). By combining of aerial laser scans with hyperspectral or multispectral imagery species can be mapped (Zhang and Qiu, 2012;Alonzo et al., 2013;Dalponte et al., 2014;Maschler et al., 2018;Marconi et al., 2019), with crown-level precision if the resolution of the sensors is sufficient (Ballanti et al., 2016;Fassnacht et al., 2016). However, these sensors are often custom-designed or prohibitively expensive where commercially available, which limits the accessibility of these surveys (Surový and Kuželka, 2019). ...
Logged forests cover four million square kilometers of the tropics, capturing carbon more rapidly than temperate forests and harboring rich biodiversity. Restoring these forests is essential to help avoid the worst impacts of climate change. Yet monitoring tropical forest recovery is challenging. We track the abundance of early-successional species in a forest restoration concession in Indonesia. If the species are carefully chosen, they can be used as an indicator of restoration progress. We present SLIC-UAV, a new pipeline for processing Unoccupied Aerial Vehicle (UAV) imagery using simple linear iterative clustering (SLIC)to map early-successional species in tropical forests. The pipeline comprises: (a) a field verified approach for manually labeling species; (b) automatic segmentation of imagery into “superpixels” and (c) machine learning classification of species based on both spectral and textural features. Creating superpixels massively reduces the dataset's dimensionality and enables the use of textural features, which improve classification accuracy. In addition, this approach is flexible with regards to the spatial distribution of training data. This allowed us to be flexible in the field and collect high-quality training data with the help of local experts. The accuracy ranged from 74.3% for a four-species classification task to 91.7% when focusing only on the key early-succesional species. We then extended these models across 100 hectares of forest, mapping species dominance and forest condition across the entire restoration project.
... Xiao et al. (2004) use spectral analysis to identify which pixels contain trees, and identify tree species based on their unique spectral properties. Alonzo et al. (2013) and Jensen et al. (2012) also applied statistical methods to various spectral properties to classify individual tree species in urban environments. Shang and Chisholm (2014) compared various traditional machine learning approaches for classification of tree species from hyperspectral imagery, and determined that these approaches outperformed statistical methods in their test region. ...
We introduce a novel deep learning method for detection of individual trees in urban environments using high-resolution multispectral aerial imagery. We use a convolutional neural network to regress a confidence map indicating the locations of individual trees, which are localized using a peak finding algorithm. Our method provides complete spatial coverage by detecting trees in both public and private spaces, and can scale to very large areas. In our study area spanning five cities in Southern California, we achieved an F-score of 0.735 and an RMSE of 2.157 m. We used our method to produce a map of all trees in the urban forest of California, indicating the potential for our method to support future urban forestry studies at unprecedented scales.
... Although hand-crafted algorithmic approaches to these problems have been developed [18], the size and novel nature of these data make data-driven machine learning approaches appealing. However, these new data sources bring with them new chal- and to a lesser extent urban [3] forests. However, with the advent of networks designed specifically to operate on pointclouds [56], new deep learning approaches to this problem that utilize raw LIDAR are appealing, but little-explored so far. ...
... The most basic method is to use spectral analysis to identify which pixels contain trees, and classify those trees' species based on their unique spectral properties; this is the approach taken by Xiao et al. [69], applied to relatively low spatial resolution (3.5m/pixel) imagery in urban Modesto, CA. More recently, Alonzo et al. [3] and Jensen et al. [27] applied statistical methods to various spectral properties to classify individual tree species in urban environments. Shang & Chisolm [59] instead compared multiple traditional machine learning techniques-support-vector machine (SVM), AdaBoost, and random forest-against traditional statistical methods, using multi-spectral imagery as input. ...
Cataloguing and classifying trees in the urban environment is a crucial step in urban and environmental planning. However, manual collection and maintenance of this data is expensive and time-consuming. Algorithmic approaches that rely on remote sensing data have been developed for tree detection in forests, though they generally struggle in the more varied urban environment. This work proposes a novel method for the detection of trees in the urban environment that applies deep learning to remote sensing data. Specifically, we train a PointNet-based neural network to predict tree locations directly from LIDAR data augmented with multi-spectral imaging. We compare this model to numerous high-performant baselines on a large and varied dataset in the Southern California region. We find that our best model outperforms all baselines with a 75.5\% F-score and 2.28 meter RMSE, while being highly efficient. We then analyze and compare the sources of errors, and how these reveal the strengths and weaknesses of each approach.
... Agarwal et al. (2013), using object-based classification of multi-spectral GeoEye imagery, were able to discriminate six vegetal species in an urban environment in Bangalore, India, with both producer´s and user´s accuracy over 80%, and kappa ≥ 0.78 for all species. Alonzo et al. (2013) used AVIRIS imagery to discriminate 15 urban tree species in Santa Barbara, California, USA. They achieved a producer´s accuracy varying from 63 to 95% and user´s accuracy varying from 46 to 99%, for an overall accuracy of 86% and kappa = 0.85. ...
Urban greenness is an element of vital importance for the population quality of life, and forest inventory is considered the most appropriate method for its assessment. Remote sensing has become an attractive alternative for the accomplishment of forest inventory, facilitating urban flora mapping. The present study aimed to identify the main species of trees in Teresina, Piauí, and evaluate the botanical identification accuracy by using high-resolution satellite images (Worldview-2) as compared to on-site inventory. We used the e-Cognition 8.7 software for the mapping, segmentation, and classification of the vegetal species and ERDAS Imagine 9.2 for accuracy verification. The NDVI (Normalized Difference Vegetation Index) was used to analyze the natural vegetation condition. The outskirts of the city presented higher values of NDVI. An amount of 1,392 individuals from 53 species and 28 families, were identified. Among these, the families Anacardiaceae (20.7%), Fabaceae (19.8%), Meliaceae (9.4%), Myrtaceae (6.9%), Arecaceae (6.1%), and Combretaceae (5.5%) were the most prevalent. Amongst the 53 species identified, the 16 most abundant were chosen for the analysis. The classification had a satisfactory result for the 16 vegetal species with a general classification accuracy of 69.43% and a kappa agreement index of 0,68. The species that obtained the highest accuracy were Ficus benjamin (87,5%), Terminalia cattapa (83,3%), Syzygium malaccense (82,4%), Mangifera indica (76,8%), Caesalpinia ferrea (75,9%), Pachira aquatica (73,9%), and Tabebuia sp (75,9%). The results showed that it is feasible, although challenging, to classify biodiverse vegetation in an urban environment using high-resolution satellite images. Our findings support the use of geotechnologies for inventorying urban forest in tropical cities.
... Despite a better spectral distinction between different broadleaf species, crown structure also appears to be the the most important discriminating factor for identifying broadleaf trees when fusing various data sources. Alonzo et al. [65], using AVIRIS imagery, concluded that the highest classification accuracies are obtained for species with large, densely foliated crowns. It is beneficial if the crown is densely foliated since this avoids contamination of background material in the spectral signature of the tree [61,65]. ...
... Alonzo et al. [65], using AVIRIS imagery, concluded that the highest classification accuracies are obtained for species with large, densely foliated crowns. It is beneficial if the crown is densely foliated since this avoids contamination of background material in the spectral signature of the tree [61,65]. Smaller tree crowns increase the risk that the pixel size of the spectral imagery is too small to avoid mixture with the background material [52]. ...
... In contrast, Alonzo et al. [18], who studied urban tree species mapping using 3.7 m AVIRIS data, found limited discriminatory value in the NIR range due to the very high within-class spectral variability in this region. The green edge, green peak and yellow edge, on the other hand, showed a larger contrast between various tree species [18,23,54,64,65]. ...
Green space is increasingly recognized as an important component of the urban environment. Adequate management and planning of urban green space is crucial to maximize its benefits for urban inhabitants and for the urban ecosystem in general. Inventorying urban vegetation is a costly and time-consuming process. The development of new remote sensing techniques to map and monitor vegetation has therefore become an important topic of interest to many scholars. Based on a comprehensive survey of the literature, this review article provides an overview of the main approaches proposed to map urban vegetation from high-resolution remotely sensed data. Studies are reviewed from three perspectives: (a) the vegetation typology, (b) the remote sensing data used and (c) the mapping approach applied. With regard to vegetation typology, a distinction is made between studies focusing on the mapping of functional vegetation types and studies performing mapping of lower-level taxonomic ranks, with the latter mainly focusing on urban trees. A wide variety of high-resolution imagery has been used by researchers for both types of mapping. The fusion of various types of remote sensing data, as well as the inclusion of phenological information through the use of multi-temporal imagery, prove to be the most promising avenues to improve mapping accuracy. With regard to mapping approaches, the use of deep learning is becoming more established, mostly for the mapping of tree species. Through this survey, several research gaps could be identified. Interest in the mapping of non-tree species in urban environments is still limited. The same holds for the mapping of understory species. Most studies focus on the mapping of public green spaces, while interest in the mapping of private green space is less common. The use of imagery with a high spatial and temporal resolution, enabling the retrieval of phenological information for mapping and monitoring vegetation at the species level, still proves to be limited in urban contexts. Hence, mapping approaches specifically tailored towards time-series analysis and the use of new data sources seem to hold great promise for advancing the field. Finally, unsupervised learning techniques and active learning, so far rarely applied in urban vegetation mapping, are also areas where significant progress can be expected.
... Very high spatial resolution (VHR) satellite images have demonstrated to be a cost-effective alternative to aerial photography for creating digital maps [12] and mapping TS [13]. Meanwhile, various hyperspectral (HS) data (e.g., Airborne Visible Infrared Imaging Spectrometer (AVIRIS) and Hyperion) have been used to classify and map TS, presenting a certain degree of success (e.g., [10,[14][15][16][17]). More recently, various light detection and ranging (LiDAR) techniques and unmanned aerial vehicle-(UAV-) based sensor techniques have been developed. ...
... The mapping accuracy (OA) reached about 90%. Similar works using airborne HS sensors' and satellite HS sensors' data ( Table 2) for TS classification have been done in [15,17,36,[73][74][75][76][77][78][79][80][81][82]. ...
... Usually, the 1st several MNFs are adopted [70,90] Canonical discriminant analysis (CDA) Search for a linear combination of independent variables to achieve maximum separation of classes (populations). Usually, the 1st 2-5 canonical variables are adopted [15,41,135] Wavelet transform (WT) ...
Timely and accurate information on tree species (TS) is crucial for developing strategies for sustainable management and conservation of artificial and natural forests. Over the last four decades, advances in remote sensing technologies have made TS classification possible. Since many studies on the topic have been conducted and their comprehensive results and novel findings have been published in the literature, it is necessary to conduct an updated review on the status, trends, potentials, and challenges and to recommend future directions. The review will provide an overview on various optical and light detection and ranging (LiDAR) sensors; present and assess current various techniques/methods for, and a general trend of method development in, TS classification; and identify limitations and recommend future directions. In this review, several concluding remarks were made. They include the following: (1) A large group of studies on the topic were using high-resolution satellite, airborne multi-/hyperspectral imagery, and airborne LiDAR data. (2) A trend of “multiple” method development for the topic was observed. (3) Machine learning methods including deep learning models were demonstrated to be significant in improving TS classification accuracy. (4) Recently, unmanned aerial vehicle- (UAV-) based sensors have caught the interest of researchers and practitioners for the topic-related research and applications. In addition, three future directions were recommended, including refining the three categories of “multiple” methods, developing novel data fusion algorithms or processing chains, and exploring new spectral unmixing algorithms to automatically extract and map TS spectral information from satellite hyperspectral data.
... Remotely-sensed satellite imagery can be used to interpolate data from field plots based upon variability of spectral signatures, estimating variation and approximating species composition across landscapes (Adelabu et al., 2013), but the spatial and temporal resolution of most satellite imagery remains a constraint (Carleer and Wolff, 2004), and higher resolution data, such as those collected from aircraft, are needed to monitor individual trees (Bergseng et al., 2015). By combining aerial laser scans with hyperspectral or multispectral imagery species can be mapped (Alonzo et al., 2013;Dalponte et al., 2014; Chapter 3 Marconi et al., 2019;Maschler et al., 2018;Zhang and Qiu, 2012), with crown-level precision if the resolution of the sensors is sufficient (Ballanti et al., 2016;Fassnacht et al., 2016). However, these sensors are often custom-designed or prohibitively expensive, where commercially available, which limits the accessibility of these surveys (Surový and Kuželka, 2019). ...
Tropical rain forests are important carbon stores and harbours of biodiversity but are being cleared at an unprecedented rate. There is an estimated 2 billion hectares of degraded forest globally, which retains a large proportion of its biodiversity. Restoration of these lands is needed to meet global commitments to combat the interlinked climate and biodiversity crises, and effective, scalable and affordable monitoring of the restoration process is essential. High resolution remote sensing technologies offer the best hope for monitoring at scale. In particular, unoccupied aerial vehicles (UAVs) offer a viable option for high spatial and temporal resolution remote sensing, though methods to guide forest restoration with these are still in their infancy. This thesis introduces approaches for the use of remote sensing data to guide tropical forest management, with particular focus on the use of UAV data in the context of restoration, looking at canopy structure, composition and dynamics. First, I introduce the context of tropical forest restoration, discussing the contribution of remote sensing to monitoring and understanding projects, with a focus on the recent developments around the use of UAVs. I also introduce the main study site of this thesis --- an ecosystem restoration concession of nearly 100 km^2 in Sumatra, Indonesia, known as Hutan Harapan. Next, I introduce a method for delineating individual tree crowns in three dimensions from remote sensing data in the form of point clouds, as created by light detection and ranging (LiDAR) and UAV structure from motion (SfM) approaches. This method, MCGC, makes use of graph cut concepts from mathematics combined with understanding of tree crown geometry and allometric scaling to automatically map tree crowns. I validate this approach using data collected in Borneo, comparing forests with three distinctive structures, showing the power of this approach to both map trees and estimate aboveground biomass. In Chapter 3, I develop a pipeline for automatic mapping of key tree species prevalence at Hutan Harapan from photographs taken from a UAV. I show it is possible to break up imagery over management units into superpixels, and through a combination of spectral and textural patterns in the imagery, train an automatic classifier to detect the species of interest from UAV imagery. I then show the power of this approach to map prevalence of key tree species indicative of the successional stage of forest recovery and demonstrate the utility of this approach for guiding management. I find that using an extra camera to take photographs with additional wavebands only slightly improved mapping accuracy. Finally, I use a combination of a LiDAR survey in 2014 and UAV surveys in 2017 and 2018 to track the effects of the strong El Niño event of 2015-16 on the canopy at Hutan Harapan, looking at 3 sites of varying recovery status spanning 100 ha of forest. I find that early-successional forest was less resistant to the drought than taller secondary forest – with canopy height loss and high mortality. However, in the subsequent high-rainfall period, I observe that early-successional forests recovered strongly. Together, the analyses demonstrate that early-successional stages lost and then regained canopy height to a greater extent that taller forest, highlighting the power of repeat surveys using LiDAR and UAVs to track canopy dynamics. Finally, I critically evaluate the methods developed, highlighting how the insights they provide can be useful for restoration practitioners, underlining the key role that remote sensing, especially with a UAV, can play whilst also needing further development.
... Research has attempted tree health detection from tree crowns using the remote sensing techniques. Hyperspectral data is useful for delineation of tree crowns with mean shift segmentation algorithm [28], pixel majority approach [29], watershed segmentation [30], forest discrimination index [31] and automatic object-based crown detection algorithms [32]. There are, however, some limitations with these techniques when delineating individual tree crowns resulting from, for example, overlapping crowns [30], defoliation and discoloration [33], variability of crown morphology, and leaf off and leaf on conditions. ...
The prevalence of black bear (Ursus americanus) bark stripping in commercial redwood (Sequoia sempervirens (D. Don) Endl.) timber stands has been increasing in recent years. This stripping is a threat to commercial timber production because of the deleterious effects on redwood tree fitness. This study sought to unveil a remote sensing method to detect these damaged trees early and map their spatial patterns. By developing a timely monitoring method, forest timber companies can manipulate their timber harvesting routines to adapt to the consequences of the problem. We explored the utility of high spatial resolution UAV-collected hyperspectral imagery as a means for early detection of individual trees stripped by black bears. A hyperspectral sensor was used to capture ultra-high spatial and spectral information pertaining to redwood trees with no damage, those that have been recently attacked by bears, and those with old bear damage. This spectral information was assessed using the Jeffries-Matusita (JM) distance to determine regions along the electromagnetic spectrum that are useful for discerning these three-health classes. While we were able to distinguish healthy trees from trees with old damage, we were unable to distinguish healthy trees from recently damaged trees due to the inherent characteristics of redwood tree growth and the subtle spectral changes within individual tree crowns for the time period assessed. The results, however, showed that with further assessment, a time window may be identified that informs damage before trees completely lose value.
... A detailed review of the effects of spectral resolution on detecting urban vegetation can be found in Fassnacht et al. (2016). Some studies using hyperspectral systems have identified important wavelength regions for classifying urban forests and trees, notably the green edge, green peak, yellow edge, red and near infrared (Xiao et al., 2004;Alonzo et al., 2013;Liu et al., 2017). Moreover, it has be argued that urban tree species can be classified using the blue region due to their relatively lower photosynthetic activity in this region (Pu and Liu, 2011). ...
... Our results showed that 8 % of papers used LiDAR to study UGSs (Supplementary Data 1: Table 2) and of these three cases focused on inventory and assessment, followed by four cases on overall UGSs mapping and four cases on three-dimensional mapping (Supplementary Data 2: Table 3). Several studies have demonstrated the benefits of combining LiDAR with hyperspectral data and high spatial resolution imagery (Zhang and Qiu, 2012;Alonzo et al., 2013;Dian et al., 2016). For instance, combination of LiDAR and hyperspectral data can aid in the detection of invasive vegetation in urban environments (Chance et al., 2016). ...
A knowledge of the characteristics of urban green spaces (UGSs) such as their abundance, spatial distribution and species composition, has an important role in a range of fields such as urban geography, urban planning and public health. Remote sensing technologies have made great contributions to the analysis of UGSs. However, a comprehensive review of the current status, challenges and potential in this area is lacking. In this paper, we scrutinize major trends in remote sensing approaches for characterising UGSs and evaluate the effectiveness of different remote sensing systems and analytical techniques. The results suggest that the number of studies focusing on mapping UGSs and classifying species within UGSs have increased rapidly over recent decades. However, there are fewer examples of non-tree species mapping, change detection, biomass and carbon mapping and vegetation health assessment within UGSs. Most studies have focused on UGSs (mainly trees) which cover large areal extents, with fewer studies of smaller patches such as street trees, urban gardens, recreational spaces and public parks, even though collectively such patches can cover substantial areas. Hence, we encourage future investigations to focus on a wider variety of different UGSs, particularly small-scale UGSs. We also recommend that research focuses on developing more effective image time series analysis techniques, methods to capture the complexity of UGSs and the use of SAR in studies of UGSs. At the same time, further research is needed to fully exploit remote sensing data within thematic applications such as monitoring changes in UGSs over time, quantifying biomass and carbon mapping and assessing vegetation health.
... Compared to color/panchromatic imagery that shows similar reflectance between different urban materials, imaging spectroscopy can measure the reflectance at narrow bands covering visible, near infrared, and short wave infrared (VSWIR) range, hence can differentiate materials with more subtle details (Herold et al., 2004). Applying high spatial and spectral resolution imagery for mapping urban vegetation and surfaces has been performed before (Alonzo et al., 2013). However, accurate mapping through classifying the acquired pixels requires a spatially fine-scale image, which is expensive to collect since it requires the airborne imaging spectrometer to fly at a low altitude. ...
Urban composition can be analyzed through spectral unmixing of images from airborne imaging spectrometers. Unmixing given a spectral library can be accomplished by set-based methods or distribution-based methods. For computational efficiency and optimal accuracy, set-based methods employ a library reduction procedure when applied to large spectral libraries. On the other hand, distribution-based methods model the library by only a few parameters, hence innately accept large libraries. A natural question arises that can distribution-based methods with the original large spectral library achieve comparable performance to set-based methods in urban imagery.
In this study, we aim to investigate the unmixing capability of several distribution-based methods, Gaussian mixture model (GMM), normal compositional model (NCM), and Beta compositional model (BCM) by comparing them to set-based methods MESMA and alternate angle minimization (AAM). The data for validation were collected by the AVIRIS sensor over the Santa Barbara region: two 16 m spatial resolution and two 4 m spatial resolution images. 64 validated regions of interest (ROI) (180 m by 180 m) were used to assess estimate accuracy. Ground truth was obtained using 1 m images leading to the following 6 classes: turfgrass, non-photosynthetic vegetation (NPV), paved, roof, soil, and tree. Spectral libraries were built by manually identifying and extracting pure spectra from both resolution images, resulting in 3287 spectra at 16 m and 15,426 spectra at 4 m. The libraries were further reduced to 61 spectra at 16 m and 95 spectra at 4 m for set-based methods. The results show that in terms of mean absolute error (MAE), GMM performed best among the distribution-based methods while MESMA performed best among the set-based methods. For 16 m data, there is no significant difference between GMM and MESMA (MAE = 0.069 vs. MAE = 0.074, p = 0.25). For 4 m data, though GMM is not as accurate as MESMA (MAE = 0.056 vs. MAE = 0.046, p = 7e − 5), it is better than AAM (MAE = 0.056 vs. MAE = 0.065, p = 0.02) which is a re-implementation of MESMA. Further evidence on a reconstructed synthetic dataset implies possible overfitting of the reduced library to the images for MESMA. These findings suggest that the distribution-based method GMM could achieve comparable unmixing accuracy to set-based methods without the need of library reduction, it may also be more stable across datasets, and the current 2-step workflow could be replaced by a single model in applying a universal spectral library.1
